MRI learning distinguishes histories of suicide attempt in bipolar depression
Source check: Source checked. Not independent scientific confirmation.
Finding: A two-level nearest-neighbor classifier used dynamic resting-state MRI features to distinguish bipolar-depression patients with a recent suicide-attempt history. After development in 176 patients, an additional 33-patient cohort yielded 72.72% accuracy. This is retrospective classification of an assessed clinical history, rather than prospective prediction of future attempts. Small external numbers, feature-selection optimism and clinical utility remain limitations.
Technology: Two-level k-nearest neighbors
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A first nearest-neighbor stage selects informative dynamic-activity voxels, and a second combines their features to classify assessed attempt histories.
- Test and result
- Patient-level classification, permutation checks and an additional 33-patient cohort with psychiatrist-assessed attempt histories.
- Limitation / still unknown
- Future-event prediction, causal brain mechanisms and clinical management benefit remain unestablished; internal supervised feature selection may be optimistic.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC10089096/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 21 Nov 2026. Review scope: Primary relevant Methods/results/limitations, exact primary front identity and current primary PubMed notice metadata; version dates distinguished..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.